Allows proving for incomplete computations

[arthurpaulino]

This PR also brings in other improvements:

• Use more basic building blocks for Evaluator::eval in order to avoid unnecessary clones. This speeds up evaluation:

eval_benchmark/eval_go_base_bls12/_10_16
                        time:   [540.20 µs 541.31 µs 542.61 µs]
                        change: [-13.016% -12.822% -12.564%] (p = 0.00 < 0.05)
                        Performance has improved.
eval_benchmark/eval_go_base_pallas/_10_16
                        time:   [539.73 µs 540.17 µs 540.57 µs]
                        change: [-14.290% -13.874% -13.618%] (p = 0.00 < 0.05)
                        Performance has improved.
eval_benchmark/eval_go_base_bls12/_10_160
                        time:   [5.1792 ms 5.1861 ms 5.1932 ms]
                        change: [-13.298% -13.154% -13.015%] (p = 0.00 < 0.05)
                        Performance has improved.
eval_benchmark/eval_go_base_pallas/_10_160
                        time:   [5.1816 ms 5.1878 ms 5.1949 ms]
                        change: [-13.680% -13.468% -13.298%] (p = 0.00 < 0.05)
                        Performance has improved.

• Fix off-by-1 errors when logging because the last frame wasn't being logged when the computation just reached the iteration limit
• Removes a \n in the logs to make them more compact vertically

Closes #526

[arthurpaulino]

These are not meta Lurk. But the feature should be added back in later when we have changeable REPL configs